hi, may be this can help you find some answers. http://www.fairchildsemi.com/ms/MS/MS-6500.pdf Regards, Mitesh On Wed, Nov 5, 2008 at 2:08 AM, Jennings, Kevin F <Kevin.Jennings@xxxxxxxxxx > wrote: > I've perused the archives and some datasheets from Micron and TI (App note > SPRAAL6A for connecting a TMS320DM643x to its DDR2 interface) and am trying > to understand what possible rationale there could be for caps on Vref. > The talk in this newsgroup regarding caps seemed to advocate caps to > provide the divider function, but to meet the +/- 2% tracking requirement on > Vref relative to Vdd/2 would then require use of 1% caps. There was also > talk of using X2Y caps to meet that requirement. There was also much talk > about how Vref is supposed to transmit the Vdd noise from the transmitter to > the receiver, seemingly oblivious to the fact that the 'transmitter' and the > 'receiver' for the data bus is itself is not fixed but depends on the memory > operation being performed. > > To my way of thinking, the generation of Vref per its defined function is > simply a case of trying to faithfully reproduce 50% of a signal 'over there' > from something that originates 'over here' where the signal in this case is > whatever Vdd noise there is 'at the transmitter'. A simple parallel > terminated transmission line would seem to be the thing to do here where the > pickup point for Vdd is the 'source/transmitter'; the processor and DDR(s) > are the loads along the line. So for a 50 ohm (as an example) PCB layer, > one would attach a 50 ohm resistor to ground after the last load just as one > would normally do to parallel terminate any old signal. To accomplish the > divide by 2 one would attach a 50 ohm resistor to Vdd (at 'some' appropriate > place). I don't typically need caps to terminate other transmission lines, > so I see no reason for why one would need one here either...am I missing > something? > > Since no one device is the 'transmitter' or the 'receiver' then one would > then have to pick some point for picking up Vdd for the division function > mentioned above. Any thoughts on what one point there to pick? At the > memory(s)? At the processor? Do you know that it actually matters? > > The other talk I see around (and is shown in the TI app note) is that all > that is needed is a resistive divider and TI says that other methods are not > recommended but then the schematic is littered with 5x .1uF caps (one from > Vdd to Vref; four from Vref to ground). They seem to be indicating use of a > cap divider in parallel with the resistive one and the three caps to ground > seem to be there to provide some form of bypass cap function. I guess the > hope is that the Vdd noise will not be in any frequency range where the 10% > tolerance caps in the divider will ever become the dominant force in the > divider. But if that hope is reality, then there wouldn't seem to be much > need for the caps in the first place when one can simply use 1% resistors > (albeit it does consume power). > > Other places have seemed to advocate caps for 'bypassing' Vref as well but > this makes no particular sense since the idea of Vref in the first place is > to transmit any noise on Vdd to each of the receivers...again am I missing > something here? Why would you want to in any way filter a signal whose > defined function is to simply be ½ of the input? > > There is also talk that Vref should be routed as a generally wider, fat > trace (the TI app note for example, mentions 20 mils, allowing for necking > down when inside the BGA array). Going back to the view that the Vref > generation is a simple transmission line problem, and that Vref is a low > current net, what rationale could there be for any extra trace width needed > for Vref? > > Any thoughts on why putting a cap anywhere near Vref would in any way > improve anything having to do with accomplishing it's required function > would be appreciated. The only thing I could come up with would possibly be > to use a small one along with the resistor to Vdd to compensate for the load > capacitance like a scope probe would...but I've never seen any mention > anywhere of anyone doing that for DDR Vref. > > Thanks in advance. > > Kevin Jennings > > > ------------------------------------------------------------------ > To unsubscribe from si-list: > si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field > > or to administer your membership from a web page, go to: > //www.freelists.org/webpage/si-list > > For help: > si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field > > > List technical documents are available at: > http://www.si-list.net > > List archives are viewable at: > //www.freelists.org/archives/si-list > or at our remote archives: > http://groups.yahoo.com/group/si-list/messages > Old (prior to June 6, 2001) list archives are viewable at: > http://www.qsl.net/wb6tpu > > > ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List technical documents are available at: http://www.si-list.net List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu